2 ms·
Re: the checker-boarding artifacts, that sounds interestingly like deconvolutional NNs treat their samples, at each deconv layer, as independent, rather than ta
by derefr 10y ago
Re: the checker-boarding artifacts, that sounds interestingly like deconvolutional NNs treat their samples, at each deconv layer, as independent, rather than taking the Rule of Addition of probability-weights (P(A ∪ B) = P(A) + P(B) - P(A ∩ B)) into account. (Would a model that did do this then be a "Bayesian Deconvolutional Neural Network"? Does that even make sense?)
I doubt this would result in audible artifacts at the audio level, though. The visual checkerboarding appears at the 1px detail-level of the output image—this would translate more to minor fluctuations along an audio sample's Nyquist frequency, which is effectively undetectable if the output sampling rate is of the usual kind (44.1kHz).